Predator naïveté has been invoked to explain the impacts of non-native predators on isolated populations that evolved with limited predation.  Such impacts have been repeatedly observed for the endangered Pahrump Poolfish, Empetrichthys latos, a desert species that evolved in isolation since the end of the Pleistocene. We tested poolfish for anti-predator responses to conspecific chemical alarm cues released from damaged epithelial tissue versus responses to distilled water as a control. Poolfish did not respond to conspecific alarm cues in terms of two well-documented components of behavioural alarm reactions: reduction in fish activity and movement out of the water column. Epidermal club cells are the presumptive source of alarm cues. We found a low density of club cells in Pahrump Poolfish relative to the well-studied Fathead Minnow, Pimephales promelas, used here as a positive control.  Therefore, anti-predator competence mediated by conspecific alarm cues does not seem to be a component of the ecology of Pahrump Poolfish. The phylogenetic context of our findings suggests that this is the first reported example of secondary loss of olfactory assessment of predation risk. These findings provide a proximate mechanism for the vulnerability of Pahrump Poolfish to non-native predators.

We conducted behavioural trials in 2017 with lab-reared F₂ generation poolfish during which activity and vertical position of poolfish was recorded before and after the introduction of conspecific alarm cue. Alarm cue was produced by euthanizing individual fish in a solution of 500 mg/L of tricaine methanesulfonate (MS-222) (NDSU Institutional Animal Care and Use Committee protocols #A15072), and filleting skin from both sides of the carcass. The fillets were laid flat on a piece of wet glass to measure skin area before transfer to a beaker of 50 mL dechlorinated tap water resting on crushed ice. For each species, the combined skin from all individuals was homogenized by a hand blender for 30s, and further diluted with distilled water to a final concentration of 1.0 cm² skin in 10 mL of water. In Fathead Minnows, 1 cm² of skin activates 58,000 L of water (Lawrence and Smith 1989, Wisenden 2008). Thus, this amount of skin extract concentrate (1 cm2 / 37.85 L) should produce a strong behavioural response in poolfish if they have a behavioural response to alarm cues. Control cue was prepared from dechlorinated tap water.  Both alarm and control cue solutions were aliquoted into 10-mL replicates and frozen at -18 ºC until needed.

Behavioural trials were conducted using 37.85-L glass aquaria under broad-spectrum fluorescent lights and maintained on a photoperiod of 12 h light: 12 h dark. Each tank contained an air-powered sponge filter with an additional 2.5 m length of airline tubing inserted into the outflow of the filter used to deliver test cues surreptitiously. A grid of 5 x 5 cm cells was drawn on the outside of the front-facing panel of each test tank.

Each test fish was acclimated for 24 h to an experimental tank and randomly assigned to either alarm cue or control treatment. Experimental fish were fed commercial flake food 60-75 min before trials began. For each trial, 50 mL of tank water was withdrawn through the delivery tube with a 60 mL syringe and discarded to rinse any residues from the delivery tube. An additional 50 mL of tank water was drawn and retained to be used later to flush test stimuli from the delivery tube into the tank.

A Canon® camcorder (model VIZIA HF R700) was placed 1.0 to 1.5 m directly across from each test tank. The camcorder recorded fish behaviour for a 5-min pre-stimulus observation period. Once completed, either control water or conspecific chemical alarm cue was introduced to the tank through the delivery tube, followed by the 50 mL flush of the previously retained tank water. Immediately after injection of stimulus, the camcorder recorded a 5-min post-stimulus observation period. Experimental tanks were drained, rinsed, and refilled with fresh water and cue injection tubes were replaced between trials.

Behavioural measures were vertical distribution and activity of the test fish, following standard operating protocols (Wisenden 2011). Vertical distribution was determined as the vertical location of test fish relative to the grid every 15 s and then averaged over each 5-min observational period. Using video recordings, activity was measured as the number of lines crossed (using the fish’s eye to determine its position) per minute during each 5-min observational period.

Post-stimulus response data were analysed using ANCOVA in JMP Pro 15 ® software (type III sums of squares, 0.05 alpha level). Treatment type (control water or alarm cue) was treated as the categorical predictor, with the pre-stimulus behaviour as a covariate.

We ran 84 trials (42 control water and 42 alarm cue). In one trial the fish displayed unusual swimming movements and in another 26 trials fish had either very low pre-stimulus activity (< 50 lines, n = 22) or very high activity (>400 lines, n=3). In such cases, responses during the post-stimulus period would be inherently limited to a one-sided response, i.e., speeding up for the slow fish, and slowing down for the fast fish.  We ran analyses both with the full data set and with a reduced data set of 58 fish (29 control trials and 29 alarm-cue trials) limited to fish with pre-stimulus movement in the range of 50-400 lines.  There was no difference in the outcomes from the two analyses sets, and thus we report the analyses based on the reduced data set (full data set analyses are available from the authors upon request).

Histological Examination

Twenty-nine Pahrump Poolfish and 7 Fathead Minnows were sacrificed using a lethal dosage of MS-222 (~500 mg/L) and a 3-4 mm section of skin was taken from the nape region (Wisenden and Smith 1998; Chivers et al. 2007). Thin sectioned histological samples were stained and mounted on slides and then digitally scanned using a MoticEasyScan Slide Scanner ® using Plan Apochromatic objective (20X/0.75) with image detail equivalent to 40X lens. Using Image-Pro Premier®, the area of epithelial tissue was calculated with smart segment tool and the number of visible club cells recorded for each sample. These data were used to estimate club cell density per mm² of skin.

Data were analysed with JMP Pro 15® software. We used a likelihood chi square to test for inter-species differences in club cell prevalence. Due to small sample sizes, we used a permutation procedure to test for differences in club cell density (club cells per mm²) where we performed a t-test to obtain the empirical difference in club cell densities between the two species and then conducted a permutation test (Manly 2007). For each random permutation, the observed club cell density estimates were randomized between the two species and the inter-species difference was calculated. This procedure was repeated 9,999 times, along with the observed empirical difference, to create a distribution of 10,000 inter-species club cell differences expected by chance. The p-value was calculated as the proportion of random inter-species differences (absolute value of the difference between means) greater than or equal to the absolute observed difference. Using the absolute value of mean difference was analogous to a two-tailed t-test.

For the behavioral dataset, we ran 84 trials (42 control water and 42 alarm cue). In one trial the fish displayed unusual swimming movements and in another 25 trials fish had either very low pre-stimulus activity (< 50 lines, n = 22) or very high activity (>400 lines, n=3). In such cases, responses during the post-stimulus period would be inherently limited to a one-sided response, i.e., speeding up for the slow fish, and slowing down for the fast fish.  We ran analyses both with the full data set and with a reduced data set of 58 fish (29 control trials and 29 alarm-cue trials) limited to fish with pre-stimulus movement in the range of 50-400 lines.